Skip to content

Merge upstream→gfx11 (redo to create merge parent link) #1377

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
eble-amd merged 108 commits into from
Oct 15, 2026
Merged

Merge upstream→gfx11 (redo to create merge parent link) #1377

Changes from all commits
Commits
Show all changes
108 commits
Select commit Hold shift + click to select a range
9dcf6bf
[BugFix] Fix DP token padding in dflash attention metadata (#56181)
TheEpicDolphin Sep 11, 2026
2d75e58
[ROCm][Kernel][DSV4] Remove tl.constexpr to avoid cold-compile churn …
amd-xavierwang Sep 11, 2026
bbaa1b9
[ROCm][CI] Extend timeout for `Basic Models (other)` (#56522)
micah-wil Sep 11, 2026
988d9b6
[Rust Frontend][gRPC] Surface engine generation errors (#56405)
alec-flowers Sep 11, 2026
9d88ceb
[KDA] Update flashKDA to support bf16 checkpoint state (#56485)
wzhao18 Sep 11, 2026
46d2b23
[ROCm][DSV4.1][Perf] Use AITER mHC for the delayed pre block (#56503)
Fangzhou-Ai Sep 12, 2026
56d7270
[Bugfix][NIXL] Don't evict a remote engine a transfer is still readin…
jyizheng Sep 12, 2026
8065045
[CI] Give every Buildkite step an explicit key, and a hook to enforce…
tahsintunan Sep 12, 2026
06e57f6
[ROCm][Kimi-K3] Fix non-contiguous state_indices crash and GPU-sync a…
divakar-amd Sep 12, 2026
a0844fa
[XPU][CI] skip DeepSeek-V4.1-Flash in `test_tensor_schema.py` (#56463)
zhenwei-intel Sep 12, 2026
31759cc
[Rust Frontend][Multimodal] Accept preprocessed multimodal gRPC featu…
biswapanda Sep 12, 2026
ec6b049
[CI] Bump CUTLASS DSL to 4.7 (#54927)
ZJY0516 Sep 12, 2026
5f3e4b7
Revert "[Agents] Link Triton skill to JIT kernel warmup guide" (#56546)
WoosukKwon Sep 12, 2026
485421b
[Rust Frontend] Honor HF revisions, offline mode, and cache directory…
BugenZhao Sep 12, 2026
120ec4e
[5/N][warmup][DSv4] Migrate NVIDIA CuTeDSL attention kernels (#53566)
LopezCastroRoberto Sep 12, 2026
d43bb2f
[3/N] HiSparse: host-resident sparse-MLA decode hot-buffering (#53781)
MatthewBonanni Sep 12, 2026
eed1f3d
[Rust Frontend] Support `generation` blocks in HF chat template (#56378)
BugenZhao Sep 12, 2026
dca96bf
[Bugfix][Spec Decode] Avoid fastsafetensors deadlock for PP draft mod…
gcanlin Sep 12, 2026
372477d
[Bugfix][Examples] Launch prefill and decode concurrently in NixlPush…
junuxyz Sep 12, 2026
30118ba
[DSV4.1] Remove compressor-aware image sentinel token padding (#56554)
Isotr0py Sep 12, 2026
d406f09
[ROCm][CI] Stage F gating (#55252)
AndreasKaratzas Sep 12, 2026
756794a
[Frontend] create unified Cohere parser (#56392)
walterbm Sep 12, 2026
33fa95a
[CI] Sample GPU utilization and memory alongside test timelines (#56541)
khluu Sep 12, 2026
6b15346
[Build] Define _USE_MATH_DEFINES for FlashMLA targets (#54007)
arcusbuilds Sep 12, 2026
d86257e
[CI/Build] Give the Torch ABI audit time to start (#56596)
khluu Sep 12, 2026
bde4feb
[XPU][CI] Remove pip install dependency in test yaml files (#55171)
jikunshang Sep 12, 2026
a0914ab
[Nano-Nemotron] Fix Nano-Nemotron precomputed multimodal embeddings (…
taneem-ibrahim Sep 12, 2026
1b29c50
[Bugfix] Initialize data parser in Nano-Nemotron audio test (#56401)
taneem-ibrahim Sep 12, 2026
986e2f8
[Refactor][ROCm] Migrate the RDNA3 W4A16 MoE to the oracle/experts pa…
JartX Sep 12, 2026
2f59050
[XPU][CI] Skip ROCm test on non-ROCm platforms (#56555)
zhenwei-intel Sep 12, 2026
13e221f
[Perf] Fuse DSV4.1 input metadata preparation with Triton (#56562)
WoosukKwon Sep 12, 2026
658c813
[Elastic EP] Reuse CUDA graphs across reconfiguration (#54985)
itayalroy Sep 12, 2026
22f6e4e
[Frontend][Rust] Reject empty structured-output values (#54821)
KernelClint Sep 12, 2026
c377114
[Bugfix][CI] Update CuTeDSL indexer Q sentinel test for migrated wrap…
LopezCastroRoberto Sep 12, 2026
9d3e991
Add @arpera to CODEOWNERS of Structured Output (#56501)
arpera Sep 12, 2026
3bb23e2
[CI/Build] Add DSv4.1 auto-label rules and narrow DSv4 (#56332)
jcotant-inferact Sep 12, 2026
dff76bc
[Bugfix][CI] skip conftest for NPU compatibility test (#55799)
yzeyu71 Sep 12, 2026
29332cf
[4/N] Expose HiSparse cache metrics via KV connector stats (#56061)
NickLucche Sep 12, 2026
7bcca16
[Bugfix] Fix DeepGEMM FP8 warmup coverage (#56452)
ayush1399 Sep 12, 2026
fd7cd4b
[CI] Fix Qwen3 Omni DSpark load test config (#56600)
khluu Sep 12, 2026
1ee4be4
[CI] Update DeepSeek V4.1 MegaMoE routing test (#56599)
khluu Sep 12, 2026
7ee8a6d
[watermarking] Dual-key gumbel-max watermarking for speculative decod…
TQCB Sep 12, 2026
72d4d83
[ROCm][Bugfix] Fix elastic EP scaling deadlock (#56610)
mawong-amd Sep 12, 2026
ebe1dec
[PCP][DCP] Enable PCP+DCP on sparse-MLA models (#56157)
PatrykSaffer Sep 12, 2026
e19a3e1
[5/N] Share HiSparse host cache across TP ranks (reopens #52760) (#56…
LucasWilkinson Sep 12, 2026
aed894c
[6/N][warmup][DSv4] Migrate sampling, and DFlash JIT kernels (#56323)
LopezCastroRoberto Sep 13, 2026
e52be1a
[9/N][warmup][DSv4] Migrate MHC TileLang kernels (#50178)
LopezCastroRoberto Sep 13, 2026
533dbe4
[Misc] Clean up fast loader daemon quant method verification (#55656)
Isotr0py Sep 13, 2026
c2ea9f1
[Kernel][Perf][Quantization] Fix odd-row performance cliff in per-tok…
ayush1399 Sep 13, 2026
c711f74
[ROCm] Auto-enable breakable CUDA graphs for DeepseekV41ForCausalLM (…
maeehart Sep 13, 2026
2671fed
[MRV2] Revert explicit Triton JIT warmup migration (#56654)
WoosukKwon Sep 13, 2026
93e051e
[Bugfix][Benchmark] Make streaming TTFT/E2E latency accounting consis…
surajm20061998 Sep 13, 2026
0cf266a
[Pooling] Support prompt embeddings in MRV2 decoder pooling (#56639)
taneem-ibrahim Sep 13, 2026
410f6da
[Bugfix][KV Offload] Submit CPU stores on no-forward steps (#56621)
LucasWilkinson Sep 13, 2026
1cfd972
[CI] Fix NIXL transfer-rank geometry fixture (#56672)
LucasWilkinson Sep 13, 2026
fa008bd
[Perf][Kernel] Integrate DeepSelect TopK for the DSA sparse indexer (…
ZJY0516 Sep 13, 2026
a987777
[Bugfix][Gemma4] Keep image kwargs out of video preprocessing (#56652)
Woolgathererer Sep 13, 2026
e7a3963
[Bugfix] Avoid repeated dummy initialization and random CPU Engram fi…
WoosukKwon Sep 13, 2026
d2d649e
[DS V4.1][Engram] Support async prefetch for offloaded engram lookups…
Juntian777 Sep 13, 2026
fa1b3b1
[Bugfix][EPD] Preserve explicit multimodal UUIDs with caches disabled…
gty111 Sep 13, 2026
c351fd3
[Core] Fix ValueError on KV load failure with a hybrid KV cache (#50388)
kebe7jun Sep 13, 2026
2e9f7bb
[Pooling] Report actual input token usage for scoring APIs (#56573)
taneem-ibrahim Sep 13, 2026
de50029
[XPU][CI] fix jit_warmup_triton_launcher (#56670)
zhenwei-intel Sep 13, 2026
dd4c841
[LoRA] Add LoRA support for DeepSeek-V4 Flash Vision (#55897)
linitra24 Sep 13, 2026
8cf9de9
[Test][Determinism] Cover VLM batch invariance in default execution m…
ShengleiFu Sep 13, 2026
82a85dc
[Attention] Add Triton/FlashInfer composite for multimodal prefix att…
MatthewBonanni Sep 13, 2026
b7e0cda
[Bugfix] Detect unloaded NVFP4 weight scales with a NaN sentinel (#52…
pavelzak Sep 13, 2026
f09c52a
[ROCm][Performance] Avoid blocking MiniMax M3 scalar upload (#56170)
Fangzhou-Ai Sep 13, 2026
dec0b5d
[Bugfix] Skip Triton autotune inspection without Triton (#56676)
khluu Sep 13, 2026
b6e2aa7
[CI] Initialize warmup registry in V2 QSA runner fixture (#56707)
LucasWilkinson Sep 13, 2026
71888f5
[Perf] Avoid redundant conversions in FP8 dummy initialization (#56688)
WoosukKwon Sep 13, 2026
7fe8fc8
[NIXL][PCP][DCP] Expose PCP producer KV shards as transfer ranks (#56…
LucasWilkinson Sep 13, 2026
586f652
[Bugfix][PCP][DCP] Respect interleave in indexer KV gather mapping (#…
LucasWilkinson Sep 13, 2026
186a1e6
[Warmup] Gemma 4 de-JITification (#55768)
LopezCastroRoberto Sep 13, 2026
b234330
[Attention] Remove DCP indexer interleave guard and test TP1 output p…
stu-cao Sep 13, 2026
319cc5e
[Performance][EPD] Reduce Python proxy serialization overhead (#56657)
gty111 Sep 13, 2026
9f03b51
[DSv4 Bug] fix dsv4 start up error `NotImplementedError: DeepSeek V4 …
yewentao256 Sep 14, 2026
3b53319
[Perf] Parallelize mHC pre-norm JIT warmup (#56683)
WoosukKwon Sep 14, 2026
fe49665
[XPU][CI] update test requirements (#56704)
zhenwei-intel Sep 14, 2026
58d45fd
[Bugfix][KV Connector] MooncakeStore: exclude non-prefix-cacheable (Q…
zhewenl Sep 14, 2026
dfa1984
[ROCm][Perf] Tune MiniMax-M3 decode top-k for short contexts (#55235)
Fangzhou-Ai Sep 14, 2026
eb42686
[Bugfix][Frontend] Support Responses text types in DeepSeek V4.1 (#56…
CedricHwong Sep 14, 2026
52dd0d7
[Fast Start] Fast loader support nnode>1 (#55468)
liusy58 Sep 14, 2026
2d9c30a
[CI][Bugfix] Complete FSE fixture R-SWA contract (#56710)
khluu Sep 14, 2026
7632f76
[Rust Frontend] Support HTTP RL weight synchronization (#56567)
zupengwang Sep 14, 2026
39545e4
Revert "[CI][XPU] Disable model runner V2 for XPU quantization test f…
chaojun-zhang Sep 14, 2026
a2685f2
[Bugfix][Multimodal] Validate base64 video payloads, matching image a…
Hotragn Sep 14, 2026
214248c
[CI] Move smaller H200 workloads to 18GB MIG slices (#56695)
khluu Sep 14, 2026
dc36fcc
[CI] Collect GPU memory telemetry for MIG slices (#56747)
khluu Sep 14, 2026
6711197
[Frontend] Move grpc_server to launchers. (#56746)
noooop Sep 14, 2026
73d2a8c
[2/2][Model Runner V2] FULL CUDA graph capture for microbatched steps…
specture724 Sep 14, 2026
10e8d56
[Bugfix][Beam Search] Respect skip_special_tokens during decoding (#5…
lxy-alexander Sep 14, 2026
cf1584f
[Refactor] Normalize DeepSeek V4.1 model package naming (#56741)
BugenZhao Sep 14, 2026
dbf49da
[Fast Start] Use GPU uuid as socket folder identifier (#56669)
Isotr0py Sep 14, 2026
663d7f6
[Bugfix] Fix --lora-modules name=path parsing when path contains '=' …
hungnnvidia Sep 14, 2026
e6b4e47
[Bugfix][Rust Frontend] Preserve selected-token logprob mode (#56406)
alec-flowers Sep 14, 2026
78e8426
[Performance] Add Triton kernel for Gemma3n sparse GELU (#48498)
BWAAEEEK Sep 14, 2026
23cfaad
[CI] Update entrypoints CI (#56763)
noooop Sep 14, 2026
b443c1c
[Perf][GLM-5.3-Flash] Use FlashKDA for KDA chunked prefill (1.7-3.8x …
JaredforReal Sep 14, 2026
dc89fdf
[CPU][Profiler] Group torch profiler tables by input shape when recor…
Chinmay-Kulkarni-AMD Sep 14, 2026
79f0be2
[Bugfix][CPU] Fix DeepSeek-R1 (FP8 MLA + MoE) correctness on CPU back…
bigPYJ1151 Sep 14, 2026
238cb2b
[Perf][GLM-5.3-Flash] Dense/masked-MHA sparse prefill for the NoPE (2…
JaredforReal Sep 14, 2026
a3f9538
Merge commit '9dcf6bf344caa7793bae0b45a7896d3f8e03a01a'
eble-amd Oct 14, 2026
12fd908
Merge commit 'd43bb2f37f87a63d3a9a299c97050443cf6a2520'
eble-amd Oct 14, 2026
844734e
Merge commit '186a1e62221c02cdf8ef0433eb4ad33d1b48c7f7'
eble-amd Oct 14, 2026
de53fa2
Merge commit 'dc89fdfb0ede2918a884faae771a49bbbe7b6cf9'
eble-amd Oct 14, 2026
8f3d9cb
Merge commit '238cb2b191630e8b02965adc772e3236d26936fe'
eble-amd Oct 14, 2026
930c88e
Merge commit '8f3d9cb8eef5'
eble-amd Oct 15, 2026
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view

These merge commits were added into this branch cleanly.

There are no new changes to show.